No description available.

Many pervasive computing applications demand expressive situational awareness, which entails an entity in the pervasive computing environment learning detailed information about its immediate and surrounding context. Much work over the past decade focused on how to acquire and represent context information. However, this work is largely egocentric, focusing on individual entities in the pervasive computing environment sensing their own context. Distributed acquisition of surrounding context information is much more challenging, largely because of the expense of communication among these resource-constrained devices. This thesis presents Grapevine, a framework for efficiently sharing context information in a localized region of a pervasive computing network, using that information to dynamically form groups defined by their shared situations, and assessing the aggregate context of that group. Grapevine’s implementation details are presented and its performance benchmarked in both simulation and live pervasive computing network deployments. / text

Dans cette thèse, nous proposons des méthodes a faible complexité pour la classification en présence d'un très grand nombre de catégories. Ces methodes permettent d'accelerer la prediction des classifieurs afin des les rendre utilisables dans les applications courantes. Nous proposons deux methodes destinées respectivement a la classification monolabel et a la classification multilabel. La première méthode utilise l'information hierarchique existante entre les catégories afin de créer un représentation binaire compact de celles-ci. La seconde approche , destinée aux problemes multilabel adpate le framework des Filtres de Bloom a la representation de sous ensembles de labels sous forme de de vecteurs binaires sparses. Dans chacun des cas, des classifieurs binaires sont appris afin de prédire les representations des catégories/labels et un algorithme permettant de retrouver l'ensemble de catégories pertinentes a partir de la représentation prédite est proposée. Les méthodes proposées sont validées par des expérience sur des données de grandes échelles et donnent des performances supérieures aux méthodes classiquement utilisées pour la classification extreme. / We propose in this thesis new methods to tackle classification problems with a large number of labes also called extreme classification. The proposed approaches aim at reducing the inference conplexity in comparison with the classical methods such as one-versus-rest in order to make learning machines usable in a real life scenario. We propose two types of methods respectively for single label and multilable classification. The first proposed approach uses existing hierarchical information among the categories in order to learn low dimensional binary representation of the categories. The second type of approaches, dedicated to multilabel problems, adapts the framework of Bloom Filters to represent subsets of labels with sparse low dimensional binary vectors. In both approaches, binary classifiers are learned to predict the new low dimensional representation of the categories and several algorithms are also proposed to recover the set of relevant labels. Large scale experiments validate the methods.

Pointer analysis is one of the key static analyses during compilation. The efficiency of several compiler optimizations and transformations depends directly on the scalability and precision of the underlying pointer analysis. Recent advances still lack an efficient and scalable context-sensitive inclusion-based pointer analysis. In this work, we propose four novel techniques to improve the scalability of context-sensitive points-to analysis for C/C++ programs. First, we develop an efficient way of storing the approximate points-to information using a multi-dimensional bloom filter (multibloom). By making use of fast hash functions and exploiting the spatial locality of the points-to information, our multibloom-based points-to analysis offers significant savings in both analysis time and memory requirement. Since the representation never resets any bit in the multibloom, no points-to information is ever lost; and the analysis is sound, though approximate. This allows a client to trade off a minimal amount of precision but gain huge savings(two orders less) in memory requirement. By making use of multiple random and independent hash functions, the algorithm also achieves high precision and runs, on an average,2×faster than Andersen’s points-to analysis. Using Mod/Ref analysis as a client, we illustrate that the precision is above 98% of that of Andersen’s analysis. Second, we devise a sound randomized algorithm that processes a group of constraints in a less precise but efficient manner and the remaining constraints in a more precise manner. By randomly choosing different groups of constraints across different runs, the analysis results in different points-to information, each of which is guaranteed to be sound. By joining the results of a few runs, the analysis obtains an approximation that is very close to the one obtained by the more precise analysis and still proves efficient in terms of the analysis time. We instantiate our technique to develop a randomized context-sensitive points-to analysis. By varying the level of randomization, a client of points-to analysis can trade off minimal precision (less than 5%) for large gain in efficiency(over 50% reduction in analysis time). We also develop an adaptive version of the randomized algorithm that carefully varies the randomization across different runs to achieve maximum benefit in terms of analysis time and precision without pre-setting the randomization percentage and the number of runs. Third, we transform the points-to analysis problem into finding a solution to a system of linear equations. By making novel use of prime factorization, we illustrate how to transform complex points-to constraints into a set of linear equations and transform the solution back as a points-to solution. We prove that our algorithm is sound and show that our technique is 1.8×faster than Andersen’s analysis for large benchmarks. Finally, we observe that the order in which points-to constraints are processed plays a vital role in the algorithm efficiency. We prove that finding an optimal ordering to compute the fixpoint solution is NP-Hard. We then propose a greedy heuristic based on the amount of points-to information computed by a constraint to prioritize the constraints. This results in a dynamic ordering of the constraint evaluation which, in turn, results in skewed evaluation of constraints where each constraint is evaluated repeatedly and different number of times in a single iteration. Our prioritized analysis achieves, on an average, an improvement of 33% over Andersen’s points-to analysis. We illustrate that our algorithms help in scaling the state-of-the-art pointer analyses. We also believe that the techniques developed would be useful for other program analyses and transformations.

Cette thèse étudie la problématique de l’indexation et de la recherche dans les tables de hachage distribuées –Distributed Hash Table (DHT). Elle propose un système de stockage distribué des résumés de documents en se basant sur leur contenu. Concrètement, la thèse utilise les Filtre de Blooms (FBs) pour représenter les résumés de documents et propose une méthode efficace d’insertion et de récupération des documents représentés par des FBs dans un index distribué sur une DHT. Le stockage basé sur contenu présente un double avantage, il permet de regrouper les documents similaires afin de les retrouver plus rapidement et en même temps, il permet de retrouver les documents en faisant des recherches par mots-clés en utilisant un FB. Cependant, la résolution d’une requête par mots-clés représentée par un filtre de Bloom constitue une opération complexe, il faut un mécanisme de localisation des filtres de Bloom de la descendance qui représentent des documents stockés dans la DHT. Ainsi, la thèse propose dans un deuxième temps, deux index de filtres de Bloom distribués sur des DHTs. Le premier système d’index proposé combine les principes d’indexation basée sur contenu et de listes inversées et répond à la problématique liée à la grande quantité de données stockée au niveau des index basés sur contenu. En effet, avec l’utilisation des filtres de Bloom de grande longueur, notre solution permet de stocker les documents sur un plus grand nombre de serveurs et de les indexer en utilisant moins d’espace. Ensuite, la thèse propose un deuxième système d’index qui supporte efficacement le traitement des requêtes de sur-ensembles (des requêtes par mots-clés) en utilisant un arbre de préfixes. Cette dernière solution exploite la distribution des données et propose une fonction de répartition paramétrable permettant d’indexer les documents avec un arbre binaire équilibré. De cette manière, les documents sont répartis efficacement sur les serveurs d’indexation. En outre, la thèse propose dans la troisième solution, une méthode efficace de localisation des documents contenant un ensemble de mots-clés donnés. Comparé aux solutions de même catégorie, cette dernière solution permet d’effectuer des recherches de sur-ensembles en un moindre coût et constitue est une base solide pour la recherche de sur-ensembles sur les systèmes d’index construits au-dessus des DHTs. Enfin, la thèse propose le prototype d’un système pair-à-pair pour l’indexation de contenus et la recherche par mots-clés. Ce prototype, prêt à être déployé dans un environnement réel, est expérimenté dans l’environnement de simulation peersim qui a permis de mesurer les performances théoriques des algorithmes développés tout au long de la thèse. / This thesis examines the problem of indexing and searching in Distributed Hash Table (DHT). It provides a distributed system for storing document summaries based on their content. Concretely, the thesis uses Bloom filters (BF) to represent document summaries and proposes an efficient method for inserting and retrieving documents represented by BFs in an index distributed on a DHT. Content-based storage has a dual advantage. It allows to group similar documents together and to find and retrieve them more quickly at the same by using Bloom filters for keywords searches. However, processing a keyword query represented by a Bloom filter is a difficult operation and requires a mechanism to locate the Bloom filters that represent documents stored in the DHT. Thus, the thesis proposes in a second time, two Bloom filters indexes schemes distributed on DHT. The first proposed index system combines the principles of content-based indexing and inverted lists and addresses the issue of the large amount of data stored by content-based indexes. Indeed, by using Bloom filters with long length, this solution allows to store documents on a large number of servers and to index them using less space. Next, the thesis proposes a second index system that efficiently supports superset queries processing (keywords-queries) using a prefix tree. This solution exploits the distribution of the data and proposes a configurable distribution function that allow to index documents with a balanced binary tree. In this way, documents are distributed efficiently on indexing servers. In addition, the thesis proposes in the third solution, an efficient method for locating documents containing a set of keywords. Compared to solutions of the same category, the latter solution makes it possible to perform subset searches at a lower cost and can be considered as a solid foundation for supersets queries processing on over-dht index systems. Finally, the thesis proposes a prototype of a peer-to-peer system for indexing content and searching by keywords. This prototype, ready to be deployed in a real environment, is experimented with peersim that allowed to measure the theoretical performances of the algorithms developed throughout the thesis.

Just as packet switched networks constituted a major breakthrough in our perception of the information exchange in computer networks so have the decoupling properties of publish/subscribe systems revolutionized the way we look at networking in the context of large scale distributed systems. The decoupling of the components of publish/subscribe systems in time, space and synchronization has created an appealing platform for the asynchronous information exchange among anonymous information producers and consumers. Moreover, the content-based nature of publish/subscribe systems provides a great degree of flexibility and expressiveness as far as construction of data flows is considered. However, a number of challenges and not yet addressed issued still exists in the area of the publish/subscribe systems. One active area of research is directed toward the problem of the efficient content delivery in the content-based publish/subscribe networks. Routing of the information based on the information itself, instead of the explicit source and destination addresses poses challenges as far as efficiency and processing times are concerned. Simultaneously, due to their decoupled nature, publish/subscribe systems introduce new challenges with respect to issues related to dependability and fail-awareness. This thesis seeks to advance the field of research in both directions. First it shows the design and implementation of routing algorithms based on the end-to-end systems design principle. Proposed routing algorithms obsolete the need to perform content-based routing within the publish/subscribe network, pushing this task to the edge of the system. Moreover, this thesis presents a fail-aware approach towards construction of the content-based publish/subscribe system along with its application to the creation of the soft state publish/subscribe system. A soft state publish/subscribe system exposes the self stabilizing behavior as far as transient timing, link and node failures are concerned. The result of this thesis is a family of the XSiena content-based publish/subscribe systems, implementing the proposed concepts and algorithms. The family of the XSiena content-based publish/subscribe systems has been a subject to rigorous evaluation, which confirms the claims made in this thesis.

Just as packet switched networks constituted a major breakthrough in our perception of the information exchange in computer networks so have the decoupling properties of publish/subscribe systems revolutionized the way we look at networking in the context of large scale distributed systems. The decoupling of the components of publish/subscribe systems in time, space and synchronization has created an appealing platform for the asynchronous information exchange among anonymous information producers and consumers. Moreover, the content-based nature of publish/subscribe systems provides a great degree of flexibility and expressiveness as far as construction of data flows is considered. However, a number of challenges and not yet addressed issued still exists in the area of the publish/subscribe systems. One active area of research is directed toward the problem of the efficient content delivery in the content-based publish/subscribe networks. Routing of the information based on the information itself, instead of the explicit source and destination addresses poses challenges as far as efficiency and processing times are concerned. Simultaneously, due to their decoupled nature, publish/subscribe systems introduce new challenges with respect to issues related to dependability and fail-awareness. This thesis seeks to advance the field of research in both directions. First it shows the design and implementation of routing algorithms based on the end-to-end systems design principle. Proposed routing algorithms obsolete the need to perform content-based routing within the publish/subscribe network, pushing this task to the edge of the system. Moreover, this thesis presents a fail-aware approach towards construction of the content-based publish/subscribe system along with its application to the creation of the soft state publish/subscribe system. A soft state publish/subscribe system exposes the self stabilizing behavior as far as transient timing, link and node failures are concerned. The result of this thesis is a family of the XSiena content-based publish/subscribe systems, implementing the proposed concepts and algorithms. The family of the XSiena content-based publish/subscribe systems has been a subject to rigorous evaluation, which confirms the claims made in this thesis.